I Said Elon Might Win the AGI Race. This Week Tesla Sent the Bill.

Back in April I wrote a post called "Hot Take: Elon Might Win the AGI Race and Nobody Is Paying Attention." The argument was not that Grok is the best model. It was that Musk is the only person in AI vertically integrating the entire stack, chips, energy, rockets, compute. My exact line was: at some point, raw compute wins.

I still think the logic holds. But this week Tesla showed everybody what that logic costs, and the market did not love the number.

Tesla reported second-quarter earnings on Wednesday. Adjusted earnings came in at 33 cents a share when Wall Street expected something in the low 50s. Revenue actually beat at $28.24 billion, up 26% from a year ago, so people are still buying the cars. The problem is what it costs to run the company now. Operating margin collapsed to 1.4%.

Free cash flow went negative for the first time since early 2024, minus $1.09 billion, because capital spending jumped 142% to $5.79 billion in a single quarter. Regulatory credits, which used to be nearly free profit, fell 67% to $146 million, per Electrek's coverage of the report.

The stock fell about 14% the next day and 18% for the week, its worst week since 2022. Musk lost roughly $130 billion on paper across Tesla and SpaceX in five trading days, and reacted the way he reacts, by posting "(Former) trillionaire" on X.

The part that actually mattered was a sentence about robots

The number people will remember from this call is not the EPS miss. It is Musk admitting, out loud, that Optimus is going to be slow.

His words: the initial part of the production curve will be "quite flat and long." He called it the hardest thing Tesla has ever tried to scale in manufacturing, because every part of the robot is new and there is no existing supply chain for most of it. This is the same guy who has talked about a million Optimus units a year, then ten million. Tesla literally ripped the Model S and Model X lines out of the Fremont factory to make room for the first Optimus production line, and the first robots off that line are not going to customers. They are going to something called the Optimus Academy, where they will practice tasks and generate training data.

I want to be fair here, because I made the bull case three months ago. None of this contradicts the April thesis. If anything it confirms the scale of the bet. Tesla guided to more than $25 billion in capital spending this year and said it is lining up debt facilities to borrow up to $30 billion more. That is exactly what "vertically integrating the entire stack" looks like when it leaves the slide deck and hits the income statement. You cannot build chip fabs, robot lines, a robotaxi fleet, and AI compute all at once and also print free cash flow. You pick one.

What I keep thinking about as a regular guy

I am not a Tesla shareholder and I am not an analyst. I am a guy who did the "spend now, payoff later" thing personally, went to culinary school, followed the passion, and learned that later does not always show up on schedule. So I have some sympathy for the shape of this story.

But there is a difference between investing in yourself and asking everyone else to fund the gap. Tesla is now openly saying it will borrow tens of billions to keep the buildout going while the core car business throws off less profit than it used to. The people covering that gap are shareholders and bondholders, and this week a meaningful chunk of them said no, or at least not at that price. Electrek put the single-day damage at more than $140 billion in market value gone.

And hovering over all of it is the merger question. A big share of the investor questions submitted for the call were about combining Tesla and SpaceX, and none of them got picked. Asked directly, Musk did not say yes and did not say no. He said there will be "more and more overlap" and that you cannot talk about combining companies on an earnings call.

Dan Ives, who has been right about Musk more often than most, puts the odds of a 2027 tie-up above 80%. If that happens, Tesla shareholders stop owning a car company with a robotics dream and start owning a slice of everything Musk has ever built, including the parts currently falling out of the sky. SpaceX stock closed the week around $115, more than 40% off its post-IPO peak.

Am I taking the April post down?

No. The compute argument is the same argument it was in April, and the Terafab logic behind it has not changed.

But I am adding a footnote in my head. "Raw compute wins" is a sentence about the destination. This earnings report was about the road, and the road is long, flat, and expensive, to borrow Musk's own framing. The next few quarters will tell us whether negative free cash flow is a toll or a lifestyle. If Tesla is still burning a billion a quarter a year from now while asking to borrow $30 billion, the question stops being whether Elon wins the AGI race. It becomes whether the people funding his entry fee run out of patience before he crosses the line.

I'll be watching the cash line instead of the rocket launches for once. That probably tells you where my head is at.